In raredd/iplotr: Interactive plotting functions

interactive plots in r

• scatter plots
• dot plots
• heat maps
• tree maps
• trend lines
• customizations
• saving iplots
• install

iplotr is a fork of Karl Broman's R/qtlcharts R package to create interactive charts aimed at QTL data for use with R/qtl.

This package reformulates the core of qtlcharts into more generic plotting functions for any type of data. The plotting functions in this package were designed to be used like base R plotting functions and use familiar syntax with similar results.

Interactive plots are particularly useful for exploratory data analysis within R but may also be included in R Markdown documents, an extension of Markdown. knitr can be used to write simply-marked-up text with chunks of R code, and output is converted to an html file for viewing in a web browser.

To include interactive plots within an R Markdown document, you just need to include the relevant calls to the R function within a code chunk as if you were creating any other type of plot.

For more information on R Markdown and knitr, see Karl's knitr in a knutshell tutorial.

# knitr::opts_chunk$set(fig.width=8, fig.height=6, message=FALSE)
library('iplotr')

scatter plots

One of the most common types of plots, scatter plots, is created with the function iscatter.

iscatter(1:5, col = 1:5, cex = 5)

The base R function plot will create a similar graph without the interactivity. Users familiar with R's base plotting functions will also be familiar with some of the parameters: col, cex, xlim/ylim, etc.

However, there are some key differences. First, col is not a vector the same length as x for coloring individual points. Rather col is a vector having length equal to the number of desired groups. Points are colored according to the group to which they are assigned.

iscatter(1:5, group = c(1, 1, 1, 2, 3), col = 4:6, cex = 5)

Another difference is the labels parameter. This argument appears in most of the interactive plots in this package and is used to label points and provide some useful information about each point.

labels can be passed as either a character vector of labels for each point or as a list of labels which will be parsed and displayed in a convenient way.

labels = list(
  ' ' = rownames(mtcars), ## use whitespace for no label ("Label: ")
  MPG = mtcars$mpg,
  'No. of gears' = mtcars$gear
)

with(mtcars, {
  iscatter(wt, mpg, group = cyl, col = 1:3, labels = labels, cex = 5,
           main = 'Motor Trend car road tests')
})

You can also use html tags in labels for better display.

v_labels <- list(
  ' ' = rownames(ivolcano),
  'log<sub>2</sub>(FC)' = round(ivolcano$logFC, 2),
  '<i>p</i>-value' = sprintf('<font color=red>%s</font>',
                             format.pval(ivolcano$pval, digits = 2, eps = 0.05))
)
with(ivolcano, {
  iscatter(logFC, -log10(pval), col = c('lightgrey','green'),
           group = pval < 0.05 & abs(logFC) > 1, labels = v_labels)
})

dot plots

Dot plots are also useful to view data interactively with added information on each point. There is no x parameter here--instead the vector of y values are grouped along the x-axis using a vector of group data. Continuous values of y may be arranged by group resembling a pseudo boxplot-style plot.

You can double-click any point to increase the size for quick identification later.

with(mtcars, idot(gear, mpg, labels = labels))

Points can further be colored using group which should be a logical vector.

with(ivolcano, {
  idot(substr(rownames(ivolcano), 1, 1), logFC,
       group = cut(pval, c(0, 0.05, 0.1, 1)),
       xlab = 'Treatment arm', labels = v_labels)
})

heat maps

icorr creates a heat map from a matrix of numeric data with rows of observations and columns of variables where each cross-section of the matrix is (optionally) linked to a scatter plot of the underlying data.

icorr(mtcars, scatterplots = FALSE)

group, col, and labels arguments are used the same as in iscatter. Additionally, the matrix can be clustered and the rows/columns reordered.

set.seed(1)
dat <- replicate(50, mtcars[, sample(1:11, 1), drop = FALSE])
dat <- do.call('cbind', dat)
icorr(dat, cluster = TRUE, group = mtcars$cyl, labels = rownames(mtcars))

tree maps

This tree map is similar to idot but also features a search bar to find specific points easily. Continuous values of y may be arranged by group resembling a pseudo boxplot-style plot.

set.seed(1)
n <- 100
ng <- 5
gr <- sort(sample(LETTERS[1:ng], n, replace = TRUE))
yv <- kinda_sort(runif(n, -1, 1), n = n / 2) * 100
itree(yv, gr, main = 'Subject response from baseline',
      ylim = c(-100,100), xlab = 'Treatment arm', ylab = '% change')

trend lines

These curves showing observations with repeated measures over time can optionally be linked to up to two interactive scatter plots. labels and group vectors are used to provide labels and color both curves and points.

As individual curves are selected and highlighted, the corresponding data points on the scatter plots are also identified. Similarly, by highlighting points on either scatter plot, the curve linked to that observation is also highlighted.

set.seed(1)
n <- 25
mm <- cbind(runif(n, 0, 10), runif(n, 10, 50), runif(n, 30, 60),
            runif(n, 40, 80), runif(n, 60, 100))
mm <- mm[order(mm[, 5]), ]
x1 <- cbind(1:n, kinda_sort(rnorm(n), n / 2))
x2 <- cbind(1:n, kinda_sort(sample(1:100, n), n / 2))

icurve(
  mat = mm, group = 1:25 %% 5 == 0,
  iscatter1 = x1, iscatter2 = x2,
  labels = list(
    Patient = 1:25,
    Disease = sample(c('NHL','HL'), 25, replace = TRUE)
  ),
  plotOpts = list(
    curves_xlab = 'Response evaluation time point',
    curves_ylab = '% response',
    scat1_xlab = 'Patient', scat1_ylab = 'Lab var 1',
    scat2_xlab = 'Patient', scat2_ylab = 'Lab var 2')
)

customizations

Each type of plot has its own set of options and customizations that affect the appearance of individual aesthetic elements. Some options such as pointcolor, pointsize, or rectcolor are common among all or most plots, but others apply only to specific plots.

See ?iplotOpts for a list of the plots in this package as well as links to the lists of options for each plot.

saving iplots

To save an interactive plot (html widget), you can use the saveWidget function from the htmlwidgets package. It is recommended to use selfcontained = TRUE (default) when saving widgets as html files--file sizes will be larger, but the plots will not depend on external resources to be displayed properly.

htmlwidgets::saveWidget(file = './scatter_plot.html',
  iscatter(rnorm(10), runif(10))
)

install

install.packages('devtools')
devtools::install_github('raredd/iplotr', build_vignettes = TRUE)

raredd/iplotr documentation built on March 19, 2021, 12:45 a.m.